CSCI5622 ML Syllabus

CSCI 5622-002 Machine Learning, Fall 2009

Tentative Syllabus

DATE		TOPIC	READ	ASSIGNMENT DUE
Mon	24-Aug	Course Goals & ML Overview	1
Wed	26-Aug	ML Overview & Experimental Design	2 slides
		SUPERVISED LEARNING
Mon	31-Aug	Decision Trees & Experimental Des	3 slides
Wed	2-Sep	ML Methods cont.	5 TM slides
Mon	7-Sep	Labor Day
Wed	9-Sep	Probability Theory & Bayesian Learning	6 slides	Problem Proposal
Mon	14-Sep	Linear regression & classification	TM ML slides
Wed	16-Sep	Perceptron and Neural Networks	4 slides	Full Proposal
Mon	21-Sep	Neural Networks cont.	4 slides	Peer Feedback (Pr Fdbk)
Wed	23-Sep	Instance-based learning	8 slides	Literature Review Plan
Mon	28-Sep	Support Vector Machines	CB SVMs NC SVMs slides
Wed	30-Sep	Kernel methods	NC Kernels Text Krnls slides	Lit Rev Notes Papers 1-2
		COMPUTATIONAL LEARNING THEORY
Mon	5-Oct	Computational Learning Theory	7 slides
Wed	7-Oct	Computational Learning Theory	7	Lit Rev Notes Papers 3-4
		ENSEMBLE METHODS
Mon	12-Oct	Ensemble methods	Bagging Boosting slides	Experiment (Exp) 1 Plan
Wed	14-Oct	Ensemble methods	Stacking Random Forests slides	Literature Review Begin Exp 1
		UNSUPERVISED LEARNING
Mon	19-Oct	Hierarchical Aglomerative clustering	ESL 14.3.12 Foundations of Statistical NLP 14.-14.1 slides	Pr Fdbk: Lit Review
Wed	21-Oct	K-means clustering	ML 6.12 ESL 14.3.6 slides
Mon	26-Oct	Mixture of Gaussians	ML 6.12 ESL 6.8 8.5 12.7
Wed	28-Oct	Snow Closure
Mon	2-Nov	Expectation Maximization	ML 6.12 DLR ESL 8.5 slides
		SEMISUPERVISED & ACTIVE LEARNING
Wed	4-Nov	Semi-Supervised Learning	Co-training Yarowsky Nigam slides
Mon	9-Nov	Semi-Supervised Learning	Coupling slides	Experiment 1 Write up
Wed	11-Nov	Active Learning	Settles
Mon	16-Nov	Active Learning	Olsson	Pr Fdbk: Exp 1 Write up
Wed	18-Nov	Online Learning	Blum
Mon	23-Nov	Fall Break
Wed	25-Nov	Fall Break
		OTHER TOPICS
Mon	30-Nov	Dimensionality Reduction	Ftr Sel Guyon PCA Ghodsi
Wed	2-Dec	Topic Modeling	Steyvers slides	Draft Paper (to Pr grp)
Mon	7-Dec	Sequence learning	HMMs Boyle HMMs Rabiner slides	Pr Fdbk: Draft Paper
Wed	9-Dec	Reinforcement Learning	13 slides Mitchell slides
		WRAP-UP TERM PROJECTS
Fri	11-Dec	No meeting		Final Paper (to me)
Wed	16-Dec 4:00PM	1777 Exposition Dr., Conf. Room 102 at entry		Class Presentations

Assignments: Please submit all assignments on the schedule to me and everyone in your peer group.

Peer Review: Please provide as much constructive feedback as possible to your peers. As the semester progresses, I anticipate that many of you will help to substantially improve your peers' papers. Peer reviews should be submitted to authors every Monday following any prior week's deadline, even if the peer feedback is not explicitly mentioned in the schedule. Only if the peer review is explicitly listed on the schedule above, submit the review to me as well as to the author. These are also the only assignments I will be reviewing in detail.

Grading your peer reviews: Only when a peer review is listed on the schedule, please gave a single grade for all of the related reviews your peers give you on a scale from 0 to 10 (e.g., at Pr Fdbk: Lit Review grade all the feedback a peer gave you related to the literature review - the plan, the paper notes, and the final review):
10=A: excellent feedback, the review(s) could have a significant impact on my project
9=A: better than expected feedback, the review(s) really helped me with my project
8=B: good feedback, the review(s) helped me fine-tune my project
7=C: okay feedback, the review(s) might be a helpful
6=D: weak feedback, the review(s) are unlikely to be helpful, but they tried
0-5=F: poor feedback, the review(s) will not really have any impact on my project

Please do not grade reviews based on spelling, grammar, tact, or flattery, (but feel free to discuss those issues with reviewers). Grade the reviews by how much you think the feedback will positively affect your learning, your research project, and the likelihood your paper will be accepted by a conference or journal, relative to how much you think a peer should be able to affect those issues. When you submit assignments, please let the reviewers know if you'd like them to focus on something specific in their review.

If you give 2 or all 3 of your reviews the same score, please rank all of them to indicate which reviews you felt were the most helpful.

Send the grades and rankings to me, no later then one week after the review's deadline. I will combine the scores from all peers and determine the final grade.

Please try to give your colleagues the best possible feedback and remember that peer reviews will account for 20% of your final grade and much of this grade will be determined by the reviewee/peer.